diff options
Diffstat (limited to 'extraconfig/tasks')
17 files changed, 440 insertions, 292 deletions
diff --git a/extraconfig/tasks/major_upgrade_block_storage.sh b/extraconfig/tasks/major_upgrade_block_storage.sh index f161c049..39861826 100644 --- a/extraconfig/tasks/major_upgrade_block_storage.sh +++ b/extraconfig/tasks/major_upgrade_block_storage.sh @@ -11,8 +11,8 @@ if [[ -n $(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "syst pushd OVS_UPGRADE echo "Attempting to downloading latest openvswitch with yumdownloader" yumdownloader --resolve openvswitch - echo "Updating openvswitch with nopostun" - rpm -U --nopostun ./*.rpm + echo "Updating openvswitch with nopostun option" + rpm -U --replacepkgs --nopostun ./*.rpm popd else echo "Skipping manual upgrade of openvswitch - no restart in postun detected" diff --git a/extraconfig/tasks/major_upgrade_ceph_storage.sh b/extraconfig/tasks/major_upgrade_ceph_storage.sh index e690a383..d84cad45 100644 --- a/extraconfig/tasks/major_upgrade_ceph_storage.sh +++ b/extraconfig/tasks/major_upgrade_ceph_storage.sh @@ -50,14 +50,14 @@ timeout 60 bash -c "while kill -0 ${OSD_PIDS} 2> /dev/null; do done" # Special-case OVS for https://bugs.launchpad.net/tripleo/+bug/1635205 -if [[ -n \$(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "systemctl.*try-restart") ]]; then +if [[ -n $(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "systemctl.*try-restart") ]]; then echo "Manual upgrade of openvswitch - restart in postun detected" mkdir OVS_UPGRADE || true pushd OVS_UPGRADE echo "Attempting to downloading latest openvswitch with yumdownloader" yumdownloader --resolve openvswitch - echo "Updating openvswitch with nopostun" - rpm -U --nopostun ./*.rpm + echo "Updating openvswitch with nopostun option" + rpm -U --replacepkgs --nopostun ./*.rpm popd else echo "Skipping manual upgrade of openvswitch - no restart in postun detected" diff --git a/extraconfig/tasks/major_upgrade_check.sh b/extraconfig/tasks/major_upgrade_check.sh index b65f6915..8bdff5e7 100755 --- a/extraconfig/tasks/major_upgrade_check.sh +++ b/extraconfig/tasks/major_upgrade_check.sh @@ -18,14 +18,8 @@ check_pcsd() fi } -check_disk_for_mysql_dump() +mysql_need_update() { - # Where to backup current database if mysql need to be upgraded - MYSQL_BACKUP_DIR=/var/tmp/mysql_upgrade_osp - MYSQL_TEMP_UPGRADE_BACKUP_DIR=/var/lib/mysql-temp-upgrade-backup - # Spare disk ratio for extra safety - MYSQL_BACKUP_SIZE_RATIO=1.2 - # Shall we upgrade mysql data directory during the stack upgrade? if [ "$mariadb_do_major_upgrade" = "auto" ]; then ret=$(is_mysql_upgrade_needed) @@ -40,6 +34,17 @@ check_disk_for_mysql_dump() else DO_MYSQL_UPGRADE=1 fi +} + +check_disk_for_mysql_dump() +{ + # Where to backup current database if mysql need to be upgraded + MYSQL_BACKUP_DIR=/var/tmp/mysql_upgrade_osp + MYSQL_TEMP_UPGRADE_BACKUP_DIR=/var/lib/mysql-temp-upgrade-backup + # Spare disk ratio for extra safety + MYSQL_BACKUP_SIZE_RATIO=1.2 + + mysql_need_update if [ "$(hiera -c /etc/puppet/hiera.yaml bootstrap_nodeid)" = "$(facter hostname)" ]; then if [ $DO_MYSQL_UPGRADE -eq 1 ]; then diff --git a/extraconfig/tasks/major_upgrade_compute.sh b/extraconfig/tasks/major_upgrade_compute.sh index 950fe8d5..b0d42806 100644 --- a/extraconfig/tasks/major_upgrade_compute.sh +++ b/extraconfig/tasks/major_upgrade_compute.sh @@ -18,7 +18,6 @@ set -eu crudini --set /etc/nova/nova.conf upgrade_levels compute $upgrade_level_nova_compute - # Special-case OVS for https://bugs.launchpad.net/tripleo/+bug/1635205 if [[ -n \$(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "systemctl.*try-restart") ]]; then echo "Manual upgrade of openvswitch - restart in postun detected" @@ -26,8 +25,8 @@ if [[ -n \$(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "sys pushd OVS_UPGRADE echo "Attempting to downloading latest openvswitch with yumdownloader" yumdownloader --resolve openvswitch - echo "Updating openvswitch with nopostun" - rpm -U --nopostun ./*.rpm + echo "Updating openvswitch with nopostun option" + rpm -U --replacepkgs --nopostun ./*.rpm popd else echo "Skipping manual upgrade of openvswitch - no restart in postun detected" @@ -36,6 +35,10 @@ fi yum -y install python-zaqarclient # needed for os-collect-config yum -y update +# Due to bug#1640177 we need to restart compute agent +echo "Restarting openstack ceilometer agent compute" +systemctl restart openstack-ceilometer-compute + ENDOFCAT # ensure the permissions are OK diff --git a/extraconfig/tasks/major_upgrade_controller_pacemaker_1.sh b/extraconfig/tasks/major_upgrade_controller_pacemaker_1.sh index 4a10fa7e..080831ab 100755 --- a/extraconfig/tasks/major_upgrade_controller_pacemaker_1.sh +++ b/extraconfig/tasks/major_upgrade_controller_pacemaker_1.sh @@ -2,8 +2,6 @@ set -eu -cluster_sync_timeout=1800 - check_cluster check_pcsd if [[ -n $(is_bootstrap_node) ]]; then @@ -19,6 +17,11 @@ check_disk_for_mysql_dump # at the end of this script if [[ -n $(is_bootstrap_node) ]]; then STONITH_STATE=$(pcs property show stonith-enabled | grep "stonith-enabled" | awk '{ print $2 }') + # We create this empty file if stonith was set to true so we can reenable stonith in step2 + rm -f /var/tmp/stonith-true + if [ $STONITH_STATE == "true" ]; then + touch /var/tmp/stonith-true + fi pcs property set stonith-enabled=false fi @@ -28,182 +31,6 @@ fi # services will be restart as there are no other constraints if [[ -n $(is_bootstrap_node) ]]; then migrate_full_to_ng_ha - rabbitmq_mitaka_newton_upgrade -fi - -# After migrating the cluster to HA-NG the services not under pacemaker's control -# are still up and running. We need to stop them explicitely otherwise during the yum -# upgrade the rpm %post sections will try to do a systemctl try-restart <service>, which -# is going to take a long time because rabbit is down. By having the service stopped -# systemctl try-restart is a noop - -for service in $(services_to_migrate); do - manage_systemd_service stop "${service%%-clone}" - # So the reason for not reusing check_resource_systemd is that - # I have observed systemctl is-active returning unknown with at least - # one service that was stopped (See LP 1627254) - timeout=600 - tstart=$(date +%s) - tend=$(( $tstart + $timeout )) - check_interval=3 - while (( $(date +%s) < $tend )); do - if [[ "$(systemctl is-active ${service%%-clone})" = "active" ]]; then - echo "$service still active, sleeping $check_interval seconds." - sleep $check_interval - else - # we do not care if it is inactive, unknown or failed as long as it is - # not running - break - fi - - done -done - -# In case the mysql package is updated, the database on disk must be -# upgraded as well. This typically needs to happen during major -# version upgrades (e.g. 5.5 -> 5.6, 5.5 -> 10.1...) -# -# Because in-place upgrades are not supported across 2+ major versions -# (e.g. 5.5 -> 10.1), we rely on logical upgrades via dump/restore cycle -# https://bugzilla.redhat.com/show_bug.cgi?id=1341968 -# -# The default is to determine automatically if upgrade is needed based -# on mysql package versionning, but this can be overriden manually -# to support specific upgrade scenario - -if [[ -n $(is_bootstrap_node) ]]; then - if [ $DO_MYSQL_UPGRADE -eq 1 ]; then - mysqldump $backup_flags > "$MYSQL_BACKUP_DIR/openstack_database.sql" - cp -rdp /etc/my.cnf* "$MYSQL_BACKUP_DIR" - fi - - pcs resource disable redis - check_resource redis stopped 600 - pcs resource disable rabbitmq - check_resource rabbitmq stopped 600 - pcs resource disable galera - check_resource galera stopped 600 - pcs resource disable openstack-cinder-volume - check_resource openstack-cinder-volume stopped 600 - # Disable all VIPs before stopping the cluster, so that pcs doesn't use one as a source address: - # https://bugzilla.redhat.com/show_bug.cgi?id=1330688 - for vip in $(pcs resource show | grep ocf::heartbeat:IPaddr2 | grep Started | awk '{ print $1 }'); do - pcs resource disable $vip - check_resource $vip stopped 60 - done - pcs cluster stop --all -fi - - -# Swift isn't controlled by pacemaker -systemctl_swift stop - -tstart=$(date +%s) -while systemctl is-active pacemaker; do - sleep 5 - tnow=$(date +%s) - if (( tnow-tstart > cluster_sync_timeout )) ; then - echo_error "ERROR: cluster shutdown timed out" - exit 1 - fi -done - -# The reason we do an sql dump *and* we move the old dir out of -# the way is because it gives us an extra level of safety in case -# something goes wrong during the upgrade. Once the restore is -# successful we go ahead and remove it. If the directory exists -# we bail out as it means the upgrade process had issues in the last -# run. -if [ $DO_MYSQL_UPGRADE -eq 1 ]; then - if [ -d $MYSQL_TEMP_UPGRADE_BACKUP_DIR ]; then - echo_error "ERROR: mysql backup dir already exist" - exit 1 - fi - mv /var/lib/mysql $MYSQL_TEMP_UPGRADE_BACKUP_DIR + rabbitmq_newton_ocata_upgrade fi - -# Special-case OVS for https://bugs.launchpad.net/tripleo/+bug/1635205 -if [[ -n $(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "systemctl.*try-restart") ]]; then - echo "Manual upgrade of openvswitch - restart in postun detected" - mkdir OVS_UPGRADE || true - pushd OVS_UPGRADE - echo "Attempting to downloading latest openvswitch with yumdownloader" - yumdownloader --resolve openvswitch - echo "Updating openvswitch with nopostun" - rpm -U --nopostun ./*.rpm - popd -else - echo "Skipping manual upgrade of openvswitch - no restart in postun detected" -fi - -yum -y install python-zaqarclient # needed for os-collect-config -yum -y -q update - -# We need to ensure at least those two configuration settings, otherwise -# mariadb 10.1+ won't activate galera replication. -# wsrep_cluster_address must only be set though, its value does not -# matter because it's overriden by the galera resource agent. -cat >> /etc/my.cnf.d/galera.cnf <<EOF -[mysqld] -wsrep_on = ON -wsrep_cluster_address = gcomm://localhost -EOF - -if [ $DO_MYSQL_UPGRADE -eq 1 ]; then - # Scripts run via heat have no HOME variable set and this confuses - # mysqladmin - export HOME=/root - - mkdir /var/lib/mysql || /bin/true - chown mysql:mysql /var/lib/mysql - chmod 0755 /var/lib/mysql - restorecon -R /var/lib/mysql/ - mysql_install_db --datadir=/var/lib/mysql --user=mysql - chown -R mysql:mysql /var/lib/mysql/ - - if [ "$(hiera -c /etc/puppet/hiera.yaml bootstrap_nodeid)" = "$(facter hostname)" ]; then - mysqld_safe --wsrep-new-cluster & - # We have a populated /root/.my.cnf with root/password here so - # we need to temporarily rename it because the newly created - # db is empty and no root password is set - mv /root/.my.cnf /root/.my.cnf.temporary - timeout 60 sh -c 'while ! mysql -e "" &> /dev/null; do sleep 1; done' - mysql -u root < "$MYSQL_BACKUP_DIR/openstack_database.sql" - mv /root/.my.cnf.temporary /root/.my.cnf - mysqladmin -u root shutdown - # The import was successful so we may remove the folder - rm -r "$MYSQL_BACKUP_DIR" - fi -fi - -# If we reached here without error we can safely blow away the origin -# mysql dir from every controller - -# TODO: What if the upgrade fails on the bootstrap node, but not on -# this controller. Data may be lost. -if [ $DO_MYSQL_UPGRADE -eq 1 ]; then - rm -r $MYSQL_TEMP_UPGRADE_BACKUP_DIR -fi - -# Let's reset the stonith back to true if it was true, before starting the cluster -if [[ -n $(is_bootstrap_node) ]]; then - if [ $STONITH_STATE == "true" ]; then - pcs -f /var/lib/pacemaker/cib/cib.xml property set stonith-enabled=true - fi -fi - -# Pin messages sent to compute nodes to kilo, these will be upgraded later -crudini --set /etc/nova/nova.conf upgrade_levels compute "$upgrade_level_nova_compute" -# https://bugzilla.redhat.com/show_bug.cgi?id=1284047 -# Change-Id: Ib3f6c12ff5471e1f017f28b16b1e6496a4a4b435 -crudini --set /etc/ceilometer/ceilometer.conf DEFAULT rpc_backend rabbit -# https://bugzilla.redhat.com/show_bug.cgi?id=1284058 -# Ifd1861e3df46fad0e44ff9b5cbd58711bbc87c97 Swift Ceilometer middleware no longer exists -crudini --set /etc/swift/proxy-server.conf pipeline:main pipeline "catch_errors healthcheck cache ratelimit tempurl formpost authtoken keystone staticweb proxy-logging proxy-server" -# LP: 1615035, required only for M/N upgrade. -crudini --set /etc/nova/nova.conf DEFAULT scheduler_host_manager host_manager -# LP: 1627450, required only for M/N upgrade -crudini --set /etc/nova/nova.conf DEFAULT scheduler_driver filter_scheduler - -crudini --set /etc/sahara/sahara.conf DEFAULT plugins ambari,cdh,mapr,vanilla,spark,storm diff --git a/extraconfig/tasks/major_upgrade_controller_pacemaker_2.sh b/extraconfig/tasks/major_upgrade_controller_pacemaker_2.sh index b3a0098c..7cc6735f 100755 --- a/extraconfig/tasks/major_upgrade_controller_pacemaker_2.sh +++ b/extraconfig/tasks/major_upgrade_controller_pacemaker_2.sh @@ -2,68 +2,186 @@ set -eu -cluster_form_timeout=600 -cluster_settle_timeout=1800 -galera_sync_timeout=600 +cluster_sync_timeout=1800 -if [[ -n $(is_bootstrap_node) ]]; then - pcs cluster start --all +# After migrating the cluster to HA-NG the services not under pacemaker's control +# are still up and running. We need to stop them explicitely otherwise during the yum +# upgrade the rpm %post sections will try to do a systemctl try-restart <service>, which +# is going to take a long time because rabbit is down. By having the service stopped +# systemctl try-restart is a noop +for service in $(services_to_migrate); do + manage_systemd_service stop "${service%%-clone}" + # So the reason for not reusing check_resource_systemd is that + # I have observed systemctl is-active returning unknown with at least + # one service that was stopped (See LP 1627254) + timeout=600 tstart=$(date +%s) - while pcs status 2>&1 | grep -E '(cluster is not currently running)|(OFFLINE:)'; do - sleep 5 - tnow=$(date +%s) - if (( tnow-tstart > cluster_form_timeout )) ; then - echo_error "ERROR: timed out forming the cluster" - exit 1 - fi + tend=$(( $tstart + $timeout )) + check_interval=3 + while (( $(date +%s) < $tend )); do + if [[ "$(systemctl is-active ${service%%-clone})" = "active" ]]; then + echo "$service still active, sleeping $check_interval seconds." + sleep $check_interval + else + # we do not care if it is inactive, unknown or failed as long as it is + # not running + break + fi + done +done - if ! timeout -k 10 $cluster_settle_timeout crm_resource --wait; then - echo_error "ERROR: timed out waiting for cluster to finish transition" - exit 1 +# In case the mysql package is updated, the database on disk must be +# upgraded as well. This typically needs to happen during major +# version upgrades (e.g. 5.5 -> 5.6, 5.5 -> 10.1...) +# +# Because in-place upgrades are not supported across 2+ major versions +# (e.g. 5.5 -> 10.1), we rely on logical upgrades via dump/restore cycle +# https://bugzilla.redhat.com/show_bug.cgi?id=1341968 +# +# The default is to determine automatically if upgrade is needed based +# on mysql package versionning, but this can be overriden manually +# to support specific upgrade scenario + +# Calling this function will set the DO_MYSQL_UPGRADE variable which is used +# later +mysql_need_update + +if [[ -n $(is_bootstrap_node) ]]; then + if [ $DO_MYSQL_UPGRADE -eq 1 ]; then + mysqldump $backup_flags > "$MYSQL_BACKUP_DIR/openstack_database.sql" + cp -rdp /etc/my.cnf* "$MYSQL_BACKUP_DIR" fi - for vip in $(pcs resource show | grep ocf::heartbeat:IPaddr2 | grep Stopped | awk '{ print $1 }'); do - pcs resource enable $vip - check_resource_pacemaker $vip started 60 + pcs resource disable redis + check_resource redis stopped 600 + pcs resource disable rabbitmq + check_resource rabbitmq stopped 600 + pcs resource disable galera + check_resource galera stopped 600 + pcs resource disable openstack-cinder-volume + check_resource openstack-cinder-volume stopped 600 + # Disable all VIPs before stopping the cluster, so that pcs doesn't use one as a source address: + # https://bugzilla.redhat.com/show_bug.cgi?id=1330688 + for vip in $(pcs resource show | grep ocf::heartbeat:IPaddr2 | grep Started | awk '{ print $1 }'); do + pcs resource disable $vip + check_resource $vip stopped 60 done + pcs cluster stop --all fi -start_or_enable_service galera -check_resource galera started 600 -start_or_enable_service redis -check_resource redis started 600 -# We need mongod which is now a systemd service up and running before calling -# ceilometer-dbsync. There is still a race here: mongod might not be up on all nodes -# so ceilometer-dbsync will fail a couple of times before that. As it retries indefinitely -# we should be good. -# Due to LP Bug https://bugs.launchpad.net/tripleo/+bug/1627254 am using systemctl directly atm -systemctl start mongod -check_resource mongod started 600 -if [[ -n $(is_bootstrap_node) ]]; then - tstart=$(date +%s) - while ! clustercheck; do - sleep 5 - tnow=$(date +%s) - if (( tnow-tstart > galera_sync_timeout )) ; then - echo_error "ERROR galera sync timed out" - exit 1 - fi - done +# Swift isn't controlled by pacemaker +systemctl_swift stop - # Run all the db syncs - # TODO: check if this can be triggered in puppet and removed from here - ceilometer-dbsync --config-file=/etc/ceilometer/ceilometer.conf - cinder-manage db sync - glance-manage --config-file=/etc/glance/glance-registry.conf db_sync - heat-manage --config-file /etc/heat/heat.conf db_sync - keystone-manage db_sync - neutron-db-manage --config-file /etc/neutron/neutron.conf --config-file /etc/neutron/plugin.ini upgrade head - nova-manage db sync - nova-manage api_db sync - nova-manage db online_data_migrations - gnocchi-upgrade - sahara-db-manage --config-file /etc/sahara/sahara.conf upgrade head +tstart=$(date +%s) +while systemctl is-active pacemaker; do + sleep 5 + tnow=$(date +%s) + if (( tnow-tstart > cluster_sync_timeout )) ; then + echo_error "ERROR: cluster shutdown timed out" + exit 1 + fi +done + +# The reason we do an sql dump *and* we move the old dir out of +# the way is because it gives us an extra level of safety in case +# something goes wrong during the upgrade. Once the restore is +# successful we go ahead and remove it. If the directory exists +# we bail out as it means the upgrade process had issues in the last +# run. +if [ $DO_MYSQL_UPGRADE -eq 1 ]; then + if [ -d $MYSQL_TEMP_UPGRADE_BACKUP_DIR ]; then + echo_error "ERROR: mysql backup dir already exist" + exit 1 + fi + mv /var/lib/mysql $MYSQL_TEMP_UPGRADE_BACKUP_DIR +fi + +# Special-case OVS for https://bugs.launchpad.net/tripleo/+bug/1635205 +if [[ -n $(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "systemctl.*try-restart") ]]; then + echo "Manual upgrade of openvswitch - restart in postun detected" + mkdir OVS_UPGRADE || true + pushd OVS_UPGRADE + echo "Attempting to downloading latest openvswitch with yumdownloader" + yumdownloader --resolve openvswitch + echo "Updating openvswitch with nopostun option" + rpm -U --replacepkgs --nopostun ./*.rpm + popd +else + echo "Skipping manual upgrade of openvswitch - no restart in postun detected" fi + +yum -y install python-zaqarclient # needed for os-collect-config +yum -y -q update + +# We need to ensure at least those two configuration settings, otherwise +# mariadb 10.1+ won't activate galera replication. +# wsrep_cluster_address must only be set though, its value does not +# matter because it's overriden by the galera resource agent. +cat >> /etc/my.cnf.d/galera.cnf <<EOF +[mysqld] +wsrep_on = ON +wsrep_cluster_address = gcomm://localhost +EOF + +if [ $DO_MYSQL_UPGRADE -eq 1 ]; then + # Scripts run via heat have no HOME variable set and this confuses + # mysqladmin + export HOME=/root + + mkdir /var/lib/mysql || /bin/true + chown mysql:mysql /var/lib/mysql + chmod 0755 /var/lib/mysql + restorecon -R /var/lib/mysql/ + mysql_install_db --datadir=/var/lib/mysql --user=mysql + chown -R mysql:mysql /var/lib/mysql/ + + if [ "$(hiera -c /etc/puppet/hiera.yaml bootstrap_nodeid)" = "$(facter hostname)" ]; then + mysqld_safe --wsrep-new-cluster & + # We have a populated /root/.my.cnf with root/password here so + # we need to temporarily rename it because the newly created + # db is empty and no root password is set + mv /root/.my.cnf /root/.my.cnf.temporary + timeout 60 sh -c 'while ! mysql -e "" &> /dev/null; do sleep 1; done' + mysql -u root < "$MYSQL_BACKUP_DIR/openstack_database.sql" + mv /root/.my.cnf.temporary /root/.my.cnf + mysqladmin -u root shutdown + # The import was successful so we may remove the folder + rm -r "$MYSQL_BACKUP_DIR" + fi +fi + +# If we reached here without error we can safely blow away the origin +# mysql dir from every controller + +# TODO: What if the upgrade fails on the bootstrap node, but not on +# this controller. Data may be lost. +if [ $DO_MYSQL_UPGRADE -eq 1 ]; then + rm -r $MYSQL_TEMP_UPGRADE_BACKUP_DIR +fi + +# Let's reset the stonith back to true if it was true, before starting the cluster +if [[ -n $(is_bootstrap_node) ]]; then + if [ -f /var/tmp/stonith-true ]; then + pcs -f /var/lib/pacemaker/cib/cib.xml property set stonith-enabled=true + fi + rm -f /var/tmp/stonith-true +fi + +# Pin messages sent to compute nodes to kilo, these will be upgraded later +crudini --set /etc/nova/nova.conf upgrade_levels compute "$upgrade_level_nova_compute" +# https://bugzilla.redhat.com/show_bug.cgi?id=1284047 +# Change-Id: Ib3f6c12ff5471e1f017f28b16b1e6496a4a4b435 +crudini --set /etc/ceilometer/ceilometer.conf DEFAULT rpc_backend rabbit +# https://bugzilla.redhat.com/show_bug.cgi?id=1284058 +# Ifd1861e3df46fad0e44ff9b5cbd58711bbc87c97 Swift Ceilometer middleware no longer exists +crudini --set /etc/swift/proxy-server.conf pipeline:main pipeline "catch_errors healthcheck cache ratelimit tempurl formpost authtoken keystone staticweb proxy-logging proxy-server" +# LP: 1615035, required only for M/N upgrade. +crudini --set /etc/nova/nova.conf DEFAULT scheduler_host_manager host_manager +# LP: 1627450, required only for M/N upgrade +crudini --set /etc/nova/nova.conf DEFAULT scheduler_driver filter_scheduler + +crudini --set /etc/sahara/sahara.conf DEFAULT plugins ambari,cdh,mapr,vanilla,spark,storm + diff --git a/extraconfig/tasks/major_upgrade_controller_pacemaker_3.sh b/extraconfig/tasks/major_upgrade_controller_pacemaker_3.sh index b653c7c7..6748f891 100755 --- a/extraconfig/tasks/major_upgrade_controller_pacemaker_3.sh +++ b/extraconfig/tasks/major_upgrade_controller_pacemaker_3.sh @@ -2,25 +2,67 @@ set -eu -start_or_enable_service rabbitmq -check_resource rabbitmq started 600 +cluster_form_timeout=600 +cluster_settle_timeout=1800 +galera_sync_timeout=600 + +if [[ -n $(is_bootstrap_node) ]]; then + pcs cluster start --all + + tstart=$(date +%s) + while pcs status 2>&1 | grep -E '(cluster is not currently running)|(OFFLINE:)'; do + sleep 5 + tnow=$(date +%s) + if (( tnow-tstart > cluster_form_timeout )) ; then + echo_error "ERROR: timed out forming the cluster" + exit 1 + fi + done + + if ! timeout -k 10 $cluster_settle_timeout crm_resource --wait; then + echo_error "ERROR: timed out waiting for cluster to finish transition" + exit 1 + fi + + for vip in $(pcs resource show | grep ocf::heartbeat:IPaddr2 | grep Stopped | awk '{ print $1 }'); do + pcs resource enable $vip + check_resource_pacemaker $vip started 60 + done +fi + +start_or_enable_service galera +check_resource galera started 600 start_or_enable_service redis check_resource redis started 600 -start_or_enable_service openstack-cinder-volume -check_resource openstack-cinder-volume started 600 - +# We need mongod which is now a systemd service up and running before calling +# ceilometer-dbsync. There is still a race here: mongod might not be up on all nodes +# so ceilometer-dbsync will fail a couple of times before that. As it retries indefinitely +# we should be good. +# Due to LP Bug https://bugs.launchpad.net/tripleo/+bug/1627254 am using systemctl directly atm +systemctl start mongod +check_resource mongod started 600 -# Swift isn't controled by pacemaker -systemctl_swift start +if [[ -n $(is_bootstrap_node) ]]; then + tstart=$(date +%s) + while ! clustercheck; do + sleep 5 + tnow=$(date +%s) + if (( tnow-tstart > galera_sync_timeout )) ; then + echo_error "ERROR galera sync timed out" + exit 1 + fi + done -# We need to start the systemd services we explicitely stopped at step _1.sh -# FIXME: Should we let puppet during the convergence step do the service enabling or -# should we add it here? -services=$(services_to_migrate) -if [[ ${keep_sahara_services_on_upgrade} =~ [Ff]alse ]] ; then - services=${services%%openstack-sahara*} + # Run all the db syncs + # TODO: check if this can be triggered in puppet and removed from here + ceilometer-upgrade --config-file=/etc/ceilometer/ceilometer.conf --skip-gnocchi-resource-types + cinder-manage db sync + glance-manage --config-file=/etc/glance/glance-registry.conf db_sync + heat-manage --config-file /etc/heat/heat.conf db_sync + keystone-manage db_sync + neutron-db-manage --config-file /etc/neutron/neutron.conf --config-file /etc/neutron/plugin.ini upgrade head + nova-manage db sync + nova-manage api_db sync + nova-manage db online_data_migrations + sahara-db-manage --config-file /etc/sahara/sahara.conf upgrade head fi -for service in $services; do - manage_systemd_service start "${service%%-clone}" - check_resource_systemd "${service%%-clone}" started 600 -done diff --git a/extraconfig/tasks/major_upgrade_controller_pacemaker_4.sh b/extraconfig/tasks/major_upgrade_controller_pacemaker_4.sh new file mode 100755 index 00000000..d2cb9553 --- /dev/null +++ b/extraconfig/tasks/major_upgrade_controller_pacemaker_4.sh @@ -0,0 +1,17 @@ +#!/bin/bash + +set -eu + +start_or_enable_service rabbitmq +check_resource rabbitmq started 600 +start_or_enable_service redis +check_resource redis started 600 +start_or_enable_service openstack-cinder-volume +check_resource openstack-cinder-volume started 600 + +# start httpd so keystone is available for gnocchi +# upgrade to run. +systemctl start httpd + +# Swift isn't controled by pacemaker +systemctl_swift start diff --git a/extraconfig/tasks/major_upgrade_controller_pacemaker_5.sh b/extraconfig/tasks/major_upgrade_controller_pacemaker_5.sh new file mode 100755 index 00000000..fa95f1f8 --- /dev/null +++ b/extraconfig/tasks/major_upgrade_controller_pacemaker_5.sh @@ -0,0 +1,8 @@ +#!/bin/bash + +set -eu + +if [[ -n $(is_bootstrap_node) ]]; then + # run gnocchi upgrade + gnocchi-upgrade +fi diff --git a/extraconfig/tasks/major_upgrade_controller_pacemaker_6.sh b/extraconfig/tasks/major_upgrade_controller_pacemaker_6.sh new file mode 100755 index 00000000..d569084d --- /dev/null +++ b/extraconfig/tasks/major_upgrade_controller_pacemaker_6.sh @@ -0,0 +1,15 @@ +#!/bin/bash + +set -eu + +# We need to start the systemd services we explicitely stopped at step _1.sh +# FIXME: Should we let puppet during the convergence step do the service enabling or +# should we add it here? +services=$(services_to_migrate) +if [[ ${keep_sahara_services_on_upgrade} =~ [Ff]alse ]] ; then + services=${services%%openstack-sahara*} +fi +for service in $services; do + manage_systemd_service start "${service%%-clone}" + check_resource_systemd "${service%%-clone}" started 600 +done diff --git a/extraconfig/tasks/major_upgrade_object_storage.sh b/extraconfig/tasks/major_upgrade_object_storage.sh index 750ad82c..2667bb16 100644 --- a/extraconfig/tasks/major_upgrade_object_storage.sh +++ b/extraconfig/tasks/major_upgrade_object_storage.sh @@ -30,8 +30,8 @@ if [[ -n \$(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "sys pushd OVS_UPGRADE echo "Attempting to downloading latest openvswitch with yumdownloader" yumdownloader --resolve openvswitch - echo "Updating openvswitch with nopostun" - rpm -U --nopostun ./*.rpm + echo "Updating openvswitch with nopostun option" + rpm -U --replacepkgs --nopostun ./*.rpm popd else echo "Skipping manual upgrade of openvswitch - no restart in postun detected" diff --git a/extraconfig/tasks/major_upgrade_pacemaker.yaml b/extraconfig/tasks/major_upgrade_pacemaker.yaml index 7c78d5ad..b0418a56 100644 --- a/extraconfig/tasks/major_upgrade_pacemaker.yaml +++ b/extraconfig/tasks/major_upgrade_pacemaker.yaml @@ -113,7 +113,20 @@ resources: config: list_join: - '' - - - get_file: pacemaker_common_functions.sh + - - str_replace: + template: | + #!/bin/bash + upgrade_level_nova_compute='UPGRADE_LEVEL_NOVA_COMPUTE' + params: + UPGRADE_LEVEL_NOVA_COMPUTE: {get_param: UpgradeLevelNovaCompute} + - str_replace: + template: | + #!/bin/bash + mariadb_do_major_upgrade='MYSQL_MAJOR_UPGRADE' + params: + MYSQL_MAJOR_UPGRADE: {get_param: MySqlMajorUpgrade} + - get_file: pacemaker_common_functions.sh + - get_file: major_upgrade_check.sh - get_file: major_upgrade_pacemaker_migrations.sh - get_file: major_upgrade_controller_pacemaker_2.sh @@ -132,6 +145,63 @@ resources: config: list_join: - '' + - - get_file: pacemaker_common_functions.sh + - get_file: major_upgrade_pacemaker_migrations.sh + - get_file: major_upgrade_controller_pacemaker_3.sh + + ControllerPacemakerUpgradeDeployment_Step3: + type: OS::Heat::SoftwareDeploymentGroup + depends_on: ControllerPacemakerUpgradeDeployment_Step2 + properties: + servers: {get_param: [servers, Controller]} + config: {get_resource: ControllerPacemakerUpgradeConfig_Step3} + input_values: {get_param: input_values} + + ControllerPacemakerUpgradeConfig_Step4: + type: OS::Heat::SoftwareConfig + properties: + group: script + config: + list_join: + - '' + - - get_file: pacemaker_common_functions.sh + - get_file: major_upgrade_pacemaker_migrations.sh + - get_file: major_upgrade_controller_pacemaker_4.sh + + ControllerPacemakerUpgradeDeployment_Step4: + type: OS::Heat::SoftwareDeploymentGroup + depends_on: ControllerPacemakerUpgradeDeployment_Step3 + properties: + servers: {get_param: [servers, Controller]} + config: {get_resource: ControllerPacemakerUpgradeConfig_Step4} + input_values: {get_param: input_values} + + ControllerPacemakerUpgradeConfig_Step5: + type: OS::Heat::SoftwareConfig + properties: + group: script + config: + list_join: + - '' + - - get_file: pacemaker_common_functions.sh + - get_file: major_upgrade_pacemaker_migrations.sh + - get_file: major_upgrade_controller_pacemaker_5.sh + + ControllerPacemakerUpgradeDeployment_Step5: + type: OS::Heat::SoftwareDeploymentGroup + depends_on: ControllerPacemakerUpgradeDeployment_Step4 + properties: + servers: {get_param: [servers, Controller]} + config: {get_resource: ControllerPacemakerUpgradeConfig_Step5} + input_values: {get_param: input_values} + + ControllerPacemakerUpgradeConfig_Step6: + type: OS::Heat::SoftwareConfig + properties: + group: script + config: + list_join: + - '' - - str_replace: template: | #!/bin/bash @@ -140,13 +210,12 @@ resources: KEEP_SAHARA_SERVICES_ON_UPGRADE: {get_param: KeepSaharaServicesOnUpgrade} - get_file: pacemaker_common_functions.sh - get_file: major_upgrade_pacemaker_migrations.sh - - get_file: major_upgrade_controller_pacemaker_3.sh + - get_file: major_upgrade_controller_pacemaker_6.sh - ControllerPacemakerUpgradeDeployment_Step3: + ControllerPacemakerUpgradeDeployment_Step6: type: OS::Heat::SoftwareDeploymentGroup - depends_on: ControllerPacemakerUpgradeDeployment_Step2 + depends_on: ControllerPacemakerUpgradeDeployment_Step5 properties: servers: {get_param: [servers, Controller]} - config: {get_resource: ControllerPacemakerUpgradeConfig_Step3} + config: {get_resource: ControllerPacemakerUpgradeConfig_Step6} input_values: {get_param: input_values} - diff --git a/extraconfig/tasks/major_upgrade_pacemaker_migrations.sh b/extraconfig/tasks/major_upgrade_pacemaker_migrations.sh index 7c9083a4..6d02acc8 100644 --- a/extraconfig/tasks/major_upgrade_pacemaker_migrations.sh +++ b/extraconfig/tasks/major_upgrade_pacemaker_migrations.sh @@ -179,3 +179,23 @@ function disable_standalone_ceilometer_api { fi fi } + + +# This function will make sure that the rabbitmq ha policies are converted from mitaka to newton +# In newton we had: Attributes: set_policy="ha-all ^(?!amq\.).* {"ha-mode":"all"}" +# In ocata we want: Attributes: set_policy="ha-all ^(?!amq\.).* {"ha-mode":"exactly","ha-params":2}" +# The nr "2" should be CEIL(N/2) where N is the number of Controllers (i.e. rabbit instances) +# Note that changing an attribute like this makes the rabbitmq resource restart +function rabbitmq_newton_ocata_upgrade { + if pcs resource show rabbitmq-clone | grep -q -E "Attributes:.*\"ha-mode\":\"all\""; then + # Number of controller is obtained by counting how many hostnames we + # have in controller_node_names hiera key + nr_controllers=$(($(hiera controller_node_names | grep -o "," |wc -l) + 1)) + nr_queues=$(($nr_controllers / 2 + ($nr_controllers % 2))) + if ! [ $nr_queues -gt 0 -a $nr_queues -le $nr_controllers ]; then + echo_error "ERROR: The nr. of HA queues during the M/N upgrade is out of range $nr_queues" + exit 1 + fi + pcs resource update rabbitmq set_policy='ha-all ^(?!amq\\.).* {"ha-mode":"exactly","ha-params":'"$nr_queues}" --wait=600 + fi +} diff --git a/extraconfig/tasks/mitaka_to_newton_ceilometer_wsgi_upgrade.pp b/extraconfig/tasks/mitaka_to_newton_ceilometer_wsgi_upgrade.pp index 1c376285..a8d43663 100644 --- a/extraconfig/tasks/mitaka_to_newton_ceilometer_wsgi_upgrade.pp +++ b/extraconfig/tasks/mitaka_to_newton_ceilometer_wsgi_upgrade.pp @@ -48,7 +48,13 @@ $mongodb_replset = hiera('mongodb::server::replset') $mongo_node_string = join($mongo_node_ips_with_port, ',') $database_connection = "mongodb://${mongo_node_string}/ceilometer?replicaSet=${mongodb_replset}" -include ::ceilometer +$rabbit_hosts = hiera('rabbitmq_node_ips', undef) +$rabbit_port = hiera('ceilometer::rabbit_port', 5672) +$rabbit_endpoints = suffix(any2array(normalize_ip_for_uri($rabbit_hosts)), ":${rabbit_port}") + +class { '::ceilometer' : + rabbit_hosts => $rabbit_endpoints, +} class {'::ceilometer::db': database_connection => $database_connection, diff --git a/extraconfig/tasks/pacemaker_common_functions.sh b/extraconfig/tasks/pacemaker_common_functions.sh index 4f17b69a..2c7dfc35 100755 --- a/extraconfig/tasks/pacemaker_common_functions.sh +++ b/extraconfig/tasks/pacemaker_common_functions.sh @@ -284,7 +284,7 @@ function systemctl_swift { services=$(systemctl | grep openstack-swift- | grep running | awk '{print $1}') ;; start) - enable_swift_storage=$(hiera -c /etc/puppet/hiera.yaml 'enable_swift_storage') + enable_swift_storage=$(hiera -c /etc/puppet/hiera.yaml tripleo::profile::base::swift::storage::enable_swift_storage) if [[ $enable_swift_storage != "true" ]]; then services=( openstack-swift-proxy ) fi diff --git a/extraconfig/tasks/pacemaker_resource_restart.sh b/extraconfig/tasks/pacemaker_resource_restart.sh index 3da7efec..49d39bc8 100755 --- a/extraconfig/tasks/pacemaker_resource_restart.sh +++ b/extraconfig/tasks/pacemaker_resource_restart.sh @@ -4,11 +4,14 @@ set -eux # Run if pacemaker is running, we're the bootstrap node, # and we're updating the deployment (not creating). -if [[ -n $(pcmk_running) && -n $(is_bootstrap_node) ]]; then + +RESTART_FOLDER="/var/lib/tripleo/pacemaker-restarts" + +if [[ -d "$RESTART_FOLDER" && -n $(pcmk_running) && -n $(is_bootstrap_node) ]]; then TIMEOUT=600 - SERVICES_TO_RESTART="$(ls /var/lib/tripleo/pacemaker-restarts)" PCS_STATUS_OUTPUT="$(pcs status)" + SERVICES_TO_RESTART="$(ls $RESTART_FOLDER)" for service in $SERVICES_TO_RESTART; do if ! echo "$PCS_STATUS_OUTPUT" | grep $service; then @@ -20,6 +23,11 @@ if [[ -n $(pcmk_running) && -n $(is_bootstrap_node) ]]; then for service in $SERVICES_TO_RESTART; do echo "Restarting $service..." pcs resource restart --wait=$TIMEOUT $service - rm -f /var/lib/tripleo/pacemaker-restarts/$service + rm -f "$RESTART_FOLDER"/$service done + +fi + +if [ $(systemctl is-active haproxy) = "active" ]; then + systemctl reload haproxy fi diff --git a/extraconfig/tasks/yum_update.sh b/extraconfig/tasks/yum_update.sh index 4ca0b140..8a88ee64 100755 --- a/extraconfig/tasks/yum_update.sh +++ b/extraconfig/tasks/yum_update.sh @@ -62,6 +62,19 @@ if [[ "$pacemaker_status" == "active" && \ fi fi +# Special-case OVS for https://bugs.launchpad.net/tripleo/+bug/1635205 +if [[ -n $(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "systemctl.*try-restart") ]]; then + echo "Manual upgrade of openvswitch - restart in postun detected" + mkdir OVS_UPGRADE || true + pushd OVS_UPGRADE + echo "Attempting to downloading latest openvswitch with yumdownloader" + yumdownloader --resolve openvswitch + echo "Updating openvswitch with nopostun option" + rpm -U --replacepkgs --nopostun ./*.rpm + popd +else + echo "Skipping manual upgrade of openvswitch - no restart in postun detected" +fi if [[ "$pacemaker_status" == "active" ]] ; then echo "Pacemaker running, stopping cluster node and doing full package update" @@ -73,28 +86,14 @@ if [[ "$pacemaker_status" == "active" ]] ; then pcs cluster stop fi else - echo "Upgrading openstack-puppet-modules" + echo "Upgrading openstack-puppet-modules and its dependencies" yum -q -y update openstack-puppet-modules + yum deplist openstack-puppet-modules | awk '/dependency/{print $2}' | xargs yum -q -y update echo "Upgrading other packages is handled by config management tooling" echo -n "true" > $heat_outputs_path.update_managed_packages exit 0 fi - -# Special-case OVS for https://bugs.launchpad.net/tripleo/+bug/1635205 -if [[ -n $(rpm -q --scripts openvswitch | awk '/postuninstall/,/*/' | grep "systemctl.*try-restart") ]]; then - echo "Manual upgrade of openvswitch - restart in postun detected" - mkdir OVS_UPGRADE || true - pushd OVS_UPGRADE - echo "Attempting to downloading latest openvswitch with yumdownloader" - yumdownloader --resolve openvswitch - echo "Updating openvswitch with nopostun" - rpm -U --nopostun ./*.rpm - popd -else - echo "Skipping manual upgrade of openvswitch - no restart in postun detected" -fi - command=${command:-update} full_command="yum -q -y $command $command_arguments" echo "Running: $full_command" @@ -104,6 +103,17 @@ return_code=$? echo "$result" echo "yum return code: $return_code" +# Writes any changes caused by alterations to os-net-config and bounces the +# interfaces *before* restarting the cluster. +os-net-config -c /etc/os-net-config/config.json -v --detailed-exit-codes +RETVAL=$? +if [[ $RETVAL == 2 ]]; then + echo "os-net-config: interface configuration files updated successfully" +elif [[ $RETVAL != 0 ]]; then + echo "ERROR: os-net-config configuration failed" + exit $RETVAL +fi + if [[ "$pacemaker_status" == "active" ]] ; then echo "Starting cluster node" pcs cluster start |